Google TPU 8t

Google TPU 8t is the training chip of Google’s eighth TPU generation, announced on April 22, 2026.

Vendor: Google
Architecture: 8th-gen TPU (training)
Memory: 216 GB HBM
Memory bandwidth (TB/s): 6.528 TB/s
FP4 (PFLOPS): 12.6 PFLOPS
Form factor: Cloud TPU (superpods up to 9,600 chips)
Status: Announced

Google announced TPU 8t at Google Cloud Next on April 22, 2026. Each chip has 216 GB of HBM at 6,528 GB/s and 12.6 PFLOPS of peak FP4. A TPU 8t superpod scales to 9,600 chips with two petabytes of shared HBM and 121 exaflops of FP4 compute, and runs on Google’s Axion Arm host CPUs. Google said both eighth-generation chips would be generally available later in 2026.

Memory (GB): 216 GB
Precision note (sparse or dense): Google lists peak FP4 per chip as 12.6 PFLOPS with no sparse/dense label. Bandwidth given as 6,528 GB/s.
Launch: Announced April 22, 2026 (Google Cloud Next); general availability planned later in 2026
Vendor source: cloud.google.com
Checked: October 5, 2026

Questions

When will Google TPU 8t be available?

Google announced TPU 8t on April 22, 2026 and said both eighth-generation chips would be generally available later in 2026.

How many chips are in a TPU 8t superpod?

A TPU 8t superpod scales to 9,600 chips with two petabytes of shared HBM and 121 exaflops of FP4 compute.

How much memory does TPU 8t have?

Each TPU 8t chip has 216 GB of HBM, with bandwidth given as 6,528 GB/s.

What is TPU 8t’s FP4 performance?

Google lists peak FP4 per chip as 12.6 PFLOPS, with no sparse or dense label.

Source

Every figure is taken from the vendor’s own datasheet, product page, technical documentation or announcement, linked on each chip with the date it was last checked. How we pick and check sources: editorial standards.